Ji et al. BMC Cancer 2010, 10:426 http://www.biomedcentral.com/1471-2407/10/426

RESEARCH ARTICLE Open Access Distinguishing between cancer driver and passenger alteration candidates via cross-species comparison: a pilot study Xinglai Ji1, Jie Tang1, Richard Halberg2, Dana Busam3, Steve Ferriera3, Maria Marjorette O Peña4, Chinnambally Venkataramu5, Timothy J Yeatman5, Shaying Zhao1*

Abstract Background: We are developing a cross-species comparison strategy to distinguish between cancer driver- and passenger gene alteration candidates, by utilizing the difference in genomic location of orthologous between the human and other mammals. As an initial test of this strategy, we conducted a pilot study with human colorectal cancer (CRC) and its mouse model C57BL/6J ApcMin/+, focusing on human 5q22.2 and 18q21.1-q21.2. Methods: We first performed bioinformatics analysis on the evolution of 5q22.2 and 18q21.1-q21.2 regions. Then, we performed exon-targeted sequencing, real time quantitative polymerase chain reaction (qPCR), and real time quantitative reverse transcriptase PCR (qRT-PCR) analyses on a number of genes of both regions with both human and mouse colon tumors. Results: These two regions (5q22.2 and 18q21.1-q21.2) are frequently deleted in human CRCs and encode genuine colorectal tumor suppressors APC and SMAD4. They also encode genes such as MCC (mutated in colorectal cancer) with their role in CRC etiology unknown. We have discovered that both regions are evolutionarily unstable, resulting in genes that are clustered in each human region being found scattered at several distinct loci in the genome of many other species. For instance, APC and MCC are within 200 kb apart in human 5q22.2 but are 10 Mb apart in the mouse genome. Importantly, our analyses revealed that, while known CRC driver genes APC and SMAD4 were disrupted in both human colorectal tumors and tumors from ApcMin/+ mice, the questionable MCC gene was disrupted in human tumors but appeared to be intact in mouse tumors. Conclusions: These results indicate that MCC may not actually play any causative role in early colorectal tumorigenesis. We also hypothesize that its disruption in human CRCs is likely a mere result of its close proximity to APC in the . Expanding this pilot study to the entire genome may identify more questionable genes like MCC, facilitating the discovery of new CRC driver gene candidates.

Background cellular processes and contribute to cancer initiation Cancer is a disease of the genome. As cancer initiates and progression (i.e., drivers), others emerge simply as and progresses, genomic instability develops and abnor- victims of genomic instability occurring as a result of mal genomic changes (e.g., sequence mutations, aberrant cancer progression (i.e., passengers). Clearly, finding promoter methylation, and structural lesions such as genomic abnormalities is significant, but identifying gains/losses, inversions and translocations) accumulate those that are cancer-drivers is even more meaningful. [1-6]. While some of these abnormalities disrupt normal A central aim of cancer research has been to identify driver gene alterations. This has become both urgent and increasingly challenging in recent years with the * Correspondence: [email protected] advance of sequencing and other technologies [7] as 1Department of Biochemistry and Molecular Biology, Institute of Bioinformatics, University of Georgia, Athens 30602, GA, USA well as the launch of high-throughput cancer genome Full list of author information is available at the end of the article

© 2010 Ji et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Ji et al. BMC Cancer 2010, 10:426 Page 2 of 13 http://www.biomedcentral.com/1471-2407/10/426

projects, such as the Cancer Genome Atlas (cancergen- ApcMin/+ [10,11] on genes encoded in two evolutionarily ome.nih.gov), the International Cancer Genome Consor- unstable human genomic regions 5q22.2 and 18q21.1- tium http://www.icgc.org, and others [3,4]. Researchers q21.2. have been tackling this challengebyimprovingexperi- Human CRC is one of the better understood systems mental conditions [5] and developing more sophisticated for studying the genetics of cancer initiation and pro- statistical models and functional analysis strategies gression [12-17]. The stepwise model of human CRC [3,4,6]. development and progression proposed by Vogelstein We are developing a cross-species comparison strategy and colleagues [12,13] (Figure 1) includes: 1) inactiva- that differs fundamentally from the current published tion of the APC, SMAD4,andP53 tumor suppressors; approaches described above (which study human can- 2) overactivation of the KRAS oncogene; and 3) develop- cers only). We hypothesize that driver alteration candi- ment of genomic instability in the form of either chro- dates can be distinguished from passenger candidates mosomal instability (CIN) [12-16] or microsatellite through examination of orthologous genes or genomic instability (MSI) [16,17]. loci with tumors from multiple species having the same The human 5q22.2 and 18q21.1-q21.2 regions are type of cancer. Provided that these species share similar both constantly disrupted in human CRCs [15,16], and molecular and genetic pathways of cancer development encode bona fide CRC driver genes (e.g., APC, SMAD4) and progression, abnormalities that are recurrent among as well as genes whose role in cancer etiology remains different species will have a higher probability to be dri- unclear (e.g., mutated in colorectal cancer or MCC). By vers, whereas those that are found in only one species comparison with different species, we found that both and are located in evolutionarily unstable sites will be regions are evolutionarily unstable and prone to inter- more likely to be passengers. In our studies, evolutiona- species genomic rearrangements during evolution. As a rily unstable sites are defined as regions enriched with consequence, genes clustered in each human region interspecies genomic rearrangement breakpoints [8,9]. were, by contrast, found scattered at several different However, we do acknowledge that our strategy will not loci in the genome of many other species, including the be able to identify species-specific drivers. mouse. We investigated these genes in colon tumors Our strategy rests on the same rationale that cancer from humans as well as a well-known CRC mouse researchers have been using for years: abnormalities model B6 ApcMin/+ [10,11]. We found that, while bona recurrent among different cases are more likely to be fide driver genes APC and SMAD4 were indeed dis- drivers, compared with non-recurrent events. The differ- rupted in both species (another CRC gene P53, located ence is that we are searching for events that are recur- in a different site 17p, was also altered), the questionable rent not only among different cases within the same MCC was altered in human tumors but appeared to be species, but also among different species. One significant structurally and expressionally intact in mouse tumors. advantage of our cross-species comparison strategy has Our results indicate that MCC is unlikely a player in over single-species approaches is that it utilizes the dif- early colorectal tumorigenesis, and we hypothesize that ference in the genomic location of orthologous genes its disruption in human colorectal tumors is most likely and loci, caused by interspecies genomic rearrangements a mere result of its close proximity to APC in the that occurred during evolution, for driver and passenger human genome. This study has demonstrated the pro- alteration distinction. The pilot study illustrating this is mise of using this cross-species comparative genomics described here, which compared human colorectal can- and oncology strategy for distinguishing between CRC cer (CRC) and one of its mouse models C57BL/6J (B6) driver- and passenger gene alteration candidates.

Figure 1 The human colorectal tumorigenesis model proposed by Vogelstein and colleagues [12,13] describes sequential inactivation of tumor suppressors (APC, SMAD4, and P53), activation of oncogene KRAS, and development of genomic instability. However, genes that are near the bona fide tumor suppressors and are disrupted in the human CRC appear not to be cancer-drivers, based on mouse model studies (DCC) or due to lack of evidence for the contribution to cancer (MCC). Ji et al. BMC Cancer 2010, 10:426 Page 3 of 13 http://www.biomedcentral.com/1471-2407/10/426

Methods scores from each sequence trace. Then, lucy, a sequence Materials cleaning program obtained from the J. Craig Venter Human DNA samples purified from 40 normal/tumor Institute http://www.jcvi.org/cms/research/software/, paired tissues of sporadic Dukes B tumors were pro- was used to trim off low quality bases from each end of vided by Dr. Timothy J. Yeatman from H. Lee Moffitt the sequences. Cancer Center and Research Institute in Florida. Dr. Mutation detection Yeatman has been approved for human subject use by Both forward and reverse sequences of each amplicon Institutional Review Boards of the University of South were used to assemble the final sequence for each exon. Florida. Colonic adenomas and matching normal tissues The assembled sequences from the tumor/normal paired from several B6 ApcMin/+ mice were provided by Dr. samples were aligned using the multiple sequence align- Richard Halberg of the University of Wisconsin (UW) in ment program CLUSTALW, and each alignment was Madison and Dr. Maria Marjorette O. Peña of the manually inspected to ensure its correctness. Sequence MouseCoreFacilityoftheCenterforColonCancer changes (e.g. indels, base substitutions) were identified by Research of the University of South Carolina (USC). All finding base changes between each tumor and its match- animal studies were conducted under protocols ing normal sequences. A cut off phred quality score of 20 approved by the Institutional Animal Care and Use was used to reduce false positives due to sequencing Committee of UW and USC, following the guidelines of errors, ensuring that only high quality bases with an the American Association for the Assessment and error rate of ≤ 1% were eligible for mutation findings. Accreditation of Laboratory Animal Care. Genomic DNA samples were extracted from these tissues using qRT-PCR with mouse genes the QIAGEN DNeasy Tissue kit. Primer design The primers were designed with Primer3 (frodo.wi.mit. Comparative genomics and sequence analyses edu/primer3/), with the criteria of 20 ± 1 mer primers, The genomic sequence and annotation data of human Tm of 60 ± 1°C, < 2 G or C residues in the final 5 bases, 5q22.2 and 18q21.1-q21.2 aswellastheirhomologues an amplicon size of 65-75bp, and unique sequences when from other species were downloaded from the Ensembl compared to the RefSeq-RNA database from GenBank site at http://www.ensembl.org and the University and the genomic sequences of the species involved. The of California Santa Cruz genome site at http://www.- primers were “TCTCTCCAAGCAGCGAGAAT” and genome.ucsc.edu. Sequences of different species were “ACTTGGACGCAGCTGATTCT” for APC; “AGAAGA- aligned through comparison with the BLATN program TAGACCGCCTGCAA” and “GCTGAGCTCTGACC- and/or by examining the gene order of each species. GAAGTT” for MCC; “GATCGGTGGCTCCATCCTGG” Evolutionary breakpoints were identified by examining and “GCCGGACTCATCGTACTCCTG” for b-actin; and the alignments. Repetitive sequences were identified by “AGCGAGCGACCAAAGGAACC” and “GCATGTC- using the RepeatMasker program obtained from http:// TAAGTACGCACGGC” for 18 S Ribosomal subunit. www.repeatmasker.org. The primer sequences are all indicated in the 5′-to 3′-direction. Bi-directional exon resequencing and mutation detection RNA extraction Sequencing was performed at The J. Craig Venter Insti- The total RNA was extracted using the STAT-60 tute in Maryland following the established protocols. reagent from Tel-Test (Texas), following the manufac- Briefly, primers were designed based on genomic ture’s instruction. Then, the amount of RNA was deter- sequences obtained from the Ensembl site, flanking each mined spectrophotometrically, and only samples with a exon or within the exon (for big exons) with ~400-800 260/280 ratio of ~1.8 were subjected to further analyses bp distance apart. Then, the targeted regions were described below. amplified by Hot-Start PCR and the PCR products were cDNA synthesis cleaned up using Shrimp Alkaline Phosphatase/Exonu- For each sample, 6 μg RNA was treated with DNase to clease I mix. Each PCR amplicon was sequenced from degrade any residual genomic DNA in the RNA samples both forward and reverse directions, using the Big Dye using the TURBO DNA-free™ kitfromAmbion(Texas). Terminator chemistry and the ABI 3730xl platform. Then, 1 μg RNA was used for cDNA synthesis with the Sequence trimming RETROscript kit from Ambion, following the manufac- Phred, a base call program obtained from Dr. Brent ture’s instructions. Ewing at The University of Washington http://www. qRT-PCR reactions phrap.org/phredphrapconsed.html, was used to obtain qRT-PCR reactions were performed in triplicates with the DNA base sequences and the associated quality each well containing 10 μliQ™ SYBER Green Supermix Ji et al. BMC Cancer 2010, 10:426 Page 4 of 13 http://www.biomedcentral.com/1471-2407/10/426

from Bio-Rad, 500 nM primers each, and 0.25 μlcDNA a result, even though APC and MCC are less than 200 kb synthesized above in a total reaction volume of 20 μl, apart in the human genome, they are 10 Mb apart in the with an iCycler iQ Real-Time PCR machine. The ampli- mouse genome (likely due to an inversion event occur- fication condition was: 95°C for 10 seconds, 65°C for ring in the mouse lineage [18]). For human 18q21.1- 45 seconds, and 78°C for 20 seconds for a total of q21.2 (Figure 2B), although the genes are clustered in 40 cycles. the mouse genome as in the human genome, break- points were identified both upstream (4.5 Mb from qPCR with mouse genes SMAD2) and downstream (2.5 Mb from MBD2)ofthe The same protocol described for qRT-PCR above was region [18,19]. followed except that 10 ng genomic DNA was used in Other inter-species rearrangement breakpoints, shown each well instead of cDNA. The primers flanking in Figure 2, are also located at intergenic regions, and SMAD4exon8(CDS7)were“tttcttcttagGGCCAGTT- thus have not disrupted the exon/intron structure of CAC” and “tacCAGGATGATTGGAAATGG"; and for any genes. Compared to the rest of the genome, nearly exon 4 were “CCCACTGAAGGACATTCGAT” and every breakpoint region contains significantly more “CTGTACGTCTCCGTTGATGC”.Theprimersof repetitive sequences (up to 90%). For example, the one reference controls were “TCCAGGTAGCAATGAC- between APC and MCC contains 61% of total repeats GAGA” and “GCGCAGAGCTTGTGTGTAAA” for and 26% of Alus. These percentages are significantly MBD2 exon5; “AGAACTGGTCAGTGCCTTGG”,and higherthanthoseofthewholegenome[20-22]and “TCT GCTGACTGCTGGTGTCT” for MCC exon 16; these repetitive sequences could have mediated non-alle- “AGCAGGATTGCGTCCATATC” and “CAGTTG- lic homologous recombination, facilitating inter-species GAGTTGTCGAGCAG” for MBD1 exon 10; and rearrangements [20]. “GATGTGCCTATGGTCCTGGT” and “CCTGAGC CTGTTTCGTGTCT” for KRAS exon 4. Human 5q22.2 and 18q21.1-q21.2 regions, constantly disrupted in human CRCs, encode CRC-driver genes as Mouse tumor aCGH analysis well as an apparent CRC passenger gene and another The oligo arrays were performed using the 385 K oligo- questionable gene nucleotide arrays from Roche NimbeGen, with each SMAD4 versus DCC in 18q21.1-q21.2 chip containing 385,000 oligo probes of ~50 bp unique The human 18q21.1-q21.2 region, constantly disrupted sequences selected across the mouse genome, providing in human CRC [15,16], encodes the bona fide CRC gene a resolution of 1 probe every 5 kb on average. Amplifi- SMAD4 (Figure 1), which is an essential player of the cations and deletions were identified by our newly transforming growth factor-b (TGF-b) signaling path- developed software tool SEG (manuscript submitted). way. This pathway controls a diverse set of cellular processes including cell proliferation, recognition, differ- Results entiation, and apoptosis during embryogenesis and in Human 5q22.2 and 18q21.1-q21.2 are evolutionarily mature tissues [23,24]. SMAD4 has indeed been found unstable to be deleted in human CRCs [25-27], and its driver Through comparison of the genomic sequences of the role in tumorigenesis has been demonstrated through human and other species (i.e., chimp, orangutan, rhesus mouse models by knocking out either SMAD4 alone or macaque, marmoset, mouse, rat, guinea pig, dog, cow, both APC and SMAD4 together [28,29]. opossum, platypus, chicken, lizard, xenopus, fly, worm, Another gene, DCC (deleted in colorectal cancer [30]), mosquito, honey bee, yeast, zebrafish, puffer fish, tetrao- is found to be in close proximity to SMAD4 in the don, and ciona), we found that regions 5q22.2 and human 18q21.1-q21.2 region (Figure 2). However, unlike 18q21.1-q21.2 are both evolutionarily unstable (com- SMAD4, the mouse model studies have indicated that pared with an average human genomic site). Genomic DCC is unlikely to be a CRC-driver. This is because in rearrangements were identified in many nonhuman spe- mice, inactive DCC through targeted mutagenesis [31] cies (i.e., 10 out of 14 species shown in Figure 2 and all or spontaneous mutations (Mouse Genome Informatics, other species except xenopus listed above but not http://www.informatics.jax.org) have no effect on the shown in Figure 2), within these sites and/or nearby intestine or on tumorigenesis within the intestine. regions. Hence, while genes of each region are clustered APC versus MCC in 5q22.2 in the human genome, they are scattered on two or The human 5q22.2 region encodes the best known CRC more distinct loci in the genome of many other species. tumor suppressor, APC (adenomatous polyposis coli). With several evolutionary rearrangement breakpoints The alteration of APC is the earliest event yet identified identified, the human 5q22.2 region has been divided into in human CRCs, and it is estimated that greater than several distinct loci in the mouse genome (Figure 2A). As 85% of human CRCs have somatic mutations of this Ji et al. BMC Cancer 2010, 10:426 Page 5 of 13 http://www.biomedcentral.com/1471-2407/10/426

Figure 2 Human 5q22.2 and 18q21.1-21.2 are evolutionarily unstable. A: Human 5q22.2 (112-113 Mb) is shown as the bar with its gene- coding regions (APC, MCC, etc.) shaded. When compared to species above (horse, rhesus, orangutan and chimp; with the orthologous of the human represented by the number inside the bar), no rearrangements were found. However, when compared to species below (rat, mouse, pig, dog, cow, opossum, and platypus) or nearby (chicken and lizard), rearrangements were identified within the region. Rearrangement breakpoints are indicated by gaps between the bars, with numbers inside each bar representing the Mb region of a chromosome (e.g., “18, 25.2-26.8” represents 25.2-26.8 Mb of ) or a supercontig/ultra-contig/scaffold (e.g., S1,0-3.4). “Un” stands for chromosome “Unknown” in the released genome assembly. The arrow of each bar designates the sequence direction, and a dished arrow indicates that the homology to the human extends beyond 5q22.2 shown here. B: Human 18q21.1-q21.2 (43-50 Mb) encodes three SMAD genes, two MBD genes, DCC, and a number of other genes (not shown). The same as above, no rearrangements were found when compared to species shown above the human. However, when compared to the species shown below, rearrangements were found within the region and/or nearby. In addition, many sequences are missing in the orthologous chicken and lizard sites (demonstrated by large gaps in the alignment). The question mark “?” inside or below the cow bars indicates that the human-cow alignment at this region has not been completely resolved. gene [13]. APC’s tumor-suppressor role has been con- pathway [32-35]. Other roles of APC include mediation firmed by various mouse models (see Mouse Genome of intercellular adhesion, stabilization of the cytoskele- Informatics at http://www.informatics.jax.org). APC ton, and possible regulation of the cell cycle and apopto- encodes a multifunctional whose tumor suppres- sis[36].Indeed,lossofAPC activity has a dramatic sing function is thought to come from its ability to effect on the mouse intestinal epithelium. This has been destabilize b-catenin, a key effector in the Wnt-signaling demonstrated by altered crypt/villus architecture, Ji et al. BMC Cancer 2010, 10:426 Page 6 of 13 http://www.biomedcentral.com/1471-2407/10/426

perturbed cell migration, increased cell proliferation and aphredqualityscoreof≥ 20 and a base call error rate apoptosis, as well as altered [37]. of ≤ 1%) after trimming off low quality bases. For com- Besides the bona fide tumor suppressor APC,5q22.2 parison purposes, we also performed the same sequen- also encodes the gene MCC (mutated in colorectal can- cing process on 22 pairs of human Dukes B sporadic cer [38]). However, at present no evidence indicates colorectal tumors and their matching-normal tissue whether MCC alteration contributes to human CRC samples, and obtained 8,342 successful sequences from development and progression or not. 9,360 total sequencing attempts. Thus, we achieved a Comparative studies with human and mouse tumors sequencing success rate of 94% for the mouse samples We conducted further experimental analyses focusing and 89% for the human samples. on the APC/MCC gene pair in 5q22.2; however, we did Sequence mutation identification notdosowiththeSMAD4/DCC pair in 18q21.1 for Following sequencing, we assembled the sequences for several reasons. First, unlike DCC where the mouse each exon using its forward and reverse sequences, models have somewhat excluded its role in CRC [31] based on the reference exon sequence from the pub- (see the previous section), there have been no further lished human and mouse genomes. With the assembled studies on MCC that directly or indirectly suggest or sequences, we performed multiple sequence alignments dispute its role in cancer since its alteration in human for the tumor/normal pairs using the CLUSTALW pro- CRC was first reported [38]. More importantly, we were gram. We then manually examined each alignment to conducting comparative studies between human color- ensure that the sequences were, indeed, correctly aligned ectal tumors and tumors from B6 ApcMin/+ mice. As (see Additional file 1 for the sequence alignments). shown in Figure 2, unlike SMAD4 and DCC that are We identified regions in the alignment that displayed physically close in both genomes, APC and MCC are different base(s) between the paired tumor and normal adjacent (< 200 kb) in the human genome but distant sequences. To reduce false positives due to sequencing (> 10 Mb) in the mouse genome. Thus, the APC/MCC errors, we ignored bases that have a quality score of pair provided an excellent example to showcase our below 20. Importantly, considering that missense muta- cross-species comparison strategy, utilizing the differ- tions do not necessarily change protein function as ence in the genomic location of orthologous genes in severely as nonsense or frameshift mutations, we pri- different species to distinguish between cancer driver marily focused on base changes resulting in premature and passenger alterations. stop codons and indels within coding exons. Thus, we ApcMin/+ is a well-characterized mouse model of largely ignored base changes inside noncoding exons as human CRC [10,11,39-42]. The Min (multiple intestinal well as missense mutations within coding exons. neoplasias) allele of APC was induced by ethylnitro- sourea (ENU), and carries a point mutation that gener- APC was disrupted in both human and mouse tumors, ates a premature stop codon (see the following section whereas MCC was disrupted in human tumors but intact for more details) [10,11]. Studies [41,42] indicated that in mouse tumors the remaining wild type copy of APC is lost through APC was disrupted in both human and mouse tumors somatic recombination during intestinal tumorigenesis As described previously [10,11], the Min allele of in B6 ApcMin/+ mice. Thus, tumors from this model ApcMin/+ was created by an ENU-induced T®Apoint most closely resemble the majority of human sporadic mutation at base 2,549 of the APC gene. This changes CRCs where the APC activity is lost. codon 850 from leucine-coding “UUG” to stop codon Bidirectional exon-resequencing “UAG” (mouse APC has a total of 2,843 codons with To detect sequence mutations (e.g., indels, base substi- 8,529 bases total). In our study, the base in position tutions, and small inversions), we performed bidirec- 2,549 was independently sequenced three times for tional exon-resequencing analyses on the APC and eachsampleandfoundtobeeither“T” or “A” in the MCC genes. Along with other genes in 18q21.1-q21.2 resulting sequences (Figure 3). This indicated that, (see the later sections), we were able to design primers within the same sample, some DNA molecules had “A” for 78 out of 79 total exons. With these primers we per- whereas others had “T” at this position. However, the formed PCR amplification using five pairs of colonic wild type base “T” was more dominant in normal sam- tumor (4-5 mm in size) and matching-normal tissue ples whereas the mutant base “A” was more dominant samples from two B6 ApcMin/+ mice. We then in tumor samples (based on the phred quality scores sequenced each PCR product from both directions. Out or the base call accuracy shown in Figure 3). These of 3,264 total sequencing attempts (large exons required findings were consistent with the heterozygous nature sequencing several overlapping PCR products), 3,076 of the APC locus, having both wild type and Min successful sequences were achieved. Each successful alleles in ApcMin/+ mice,aswellaswithpreviousstu- sequence had at least 100 bp high quality bases (having dies indicating that the wild type copy of APC was lost Ji et al. BMC Cancer 2010, 10:426 Page 7 of 13 http://www.biomedcentral.com/1471-2407/10/426

Figure 3 Sequence mutations of APC in the tumors from B6 ApcMin/+ mice. Top: the bases and their quality score in parenthesis, e.g. T(46), are the outputs from the base-call program phred for the base 2549 of the mouse APC gene. The three bases next to each sample (e.g., 8762- 1N), ordered by their quality score or base-call accuracy (the base call error rate equals to 10-quality score; thus a higher score means a more accurate base call), are from three separate sequences generated for the sample. Tumors 1-4 and their matching normal tissue samples were from mouse ID 8762, whereas tumor 5 and its matching normal sample were from mouse ID 8791 (thus, 8762-1N and 8762-1T are paired normal/tumor samples, and so on). Middle and bottom: The first sequence alignment is for bases 2493-2552 of mouse APC. Using the most accurate base (i.e., having the highest quality score) for base 2549 as indicated above, tumors 3-5 have a T- > A mutation (in red/green color) creating a premature stop codon. The remaining two tumors (8762-1T and 8762-2T) have a “GG” insertion at base 6448 of APC, as shown in the second sequence alignment which is for bases 6393-6450 of mouse APC. “Consensus” stands for the corresponding sequence obtained from the published mouse genome. via somatic recombination in tumors [41,42]. In the MCC was disrupted in human tumors but intact in mouse two tumors (8762-1T and 8762-2T, from the same tumors mouse) where the wild type base “T” was slightly more We sequenced each of the 17 coding exons of the dominant, we identified a “GG” insertion at base 6,448, mouse MCC gene with a total of 2,333 bases, and found also known as codon 218 (Figure 3). This causes a fra- no mutations in the mouse tumors (see Additional meshift for the rest of the protein, indicating that file 1) (our qPCR analysis with several of its exons did these two tumors did not retain their wild type APC not reveal large copy number changes either). However, allele either. Thus, besides somatic recombination, B6 stop-codons were identified in several MCC exons in ApcMin/+ mice could lose their wild type APC copy by many of the human tumors studied (see Additional acquiring additional sequence mutations. file1).Thisisconsistentwithpreviousstudiesthat Except for these two abnormalities, we did not find reported MCC being disrupted in human CRC [38]. any other stop codon or indel mutations for the mouse Thus,unlikebonafideCRCgeneAPC, MCC is dis- APC (see Additional file 1). For the human APC,how- rupted in human tumors but intact in mouse tumors in ever, we identified significantly more mutations. Stop terms of its exon sequences. codons and indels were found within coding exons Confirmation of MCC integrity in mouse tumors by qRT-PCR 2-4, 6, 8, 9, 11 and/or 15 in the 22 human tumors ana- To confirm that MCC is intact in mouse tumors, we lyzed (see Additional file 1). This is consistent with lit- performed qRT-PCR analyses to examine its expression. erature reports that APC is altered in a vast majority In order to select proper normalization controls, we of the human sporadic CRCs [13], and reveals the het- tested 4 genes (b-actin, glyceraldehyde 3-phosphate erogeneity of sporadic human cancers. The APC gene dehydrogenase or GAPDH, 18 s ribosomal subunit,and results described above demonstrated the effectiveness b-2-microglobulin), chosen based on a literature search. of our re-sequencing protocol and data-analyzing Through examination of the variation of expression in pipeline. ten pairs of mouse tumor samples and matching-normal Ji et al. BMC Cancer 2010, 10:426 Page 8 of 13 http://www.biomedcentral.com/1471-2407/10/426

tissue samples, we designated b-actin and 18 s ribosomal samples (see Additional file 1). To determine whether subunit as the normalization control genes, because no the sequencing failure of coding exon 7 in the tumors significant changes were found across the samples. We was due to large sequence deletions or technical pro- also included the APC gene in the study. blems in our sequencing procedure, we performed A total of ten pairs of adenoma and matching-normal further analyses as described below. tissue samples from several B6 ApcMin/+ mice were ana- Confirmation by qPCR lyzed. A t-test was performed to examine the differences Because qPCR is an effective strategy for detecting large in the normalized Ct values between paired adenoma deletions, we performed this analysis on a total of 10 and normal samples (Ct is the threshold cycles: the adenomas along with their matching normal samples number of cycles at which the earliest measurable fluor- from several B6 ApcMin/+ mice. We examined exon 7 escence signal can be detected in a qPCR assay; a higher (with which sequencing succeeded on most normal sam- Ct value means fewer templates). Significant statistical ples but failed on the tumors) and exon 4 (with which changes were observed for APC (t = 1.96 and p = 0.08), sequencing succeeded on all tumor/normal samples and however, not for MCC (t = 0.71 and p = 0.5). This indi- no mutation was found) of SMAD4. We examined these cates that APC was altered whereas MCC was intact in along with four reference controls including exon 16 of the adenomas at the expression level, and is consistent MCC,exon5ofMBD2,exon10ofMBD1,andexon4 with the resequencing results described above. of KRAS. These reference exons were chosen because they are all from single-copy genes in the genome and, SMAD4 was disrupted in both human and mouse tumors more importantly, no large copy number changes were We investigated the SMAD4 gene in both human and observed between tumor and normal samples. Our mouse tumors via sequencing and other methods for a sequencing (and expression in the case of MCC)ana- number of reasons. First, SMAD4 alteration contributes lyses revealed no alterations in MCC, MBD2 and MBD1 to human colorectal tumorigenesis at a relatively early exons in mouse tumors (see above). For KRAS,wedid stage, late adenoma formation (Figure 1). Because the not find any studies that reported copy number changes; mouse tumors are adenomas, we would like to know if most alterations identified were point mutations which SMAD4 is altered or intact in these tumors. This infor- would interfere less severely with the qPCR analyses, mation would provide one critical piece of evidence sup- compared with large amplifications/deletions. porting whether or not tumorigenesis of ApcMin/+ in We noted that 8 of the 10 total tumors had a higher mice follows similar pathways as in humans (Figure 1). Ct value than their matching normal samples for exon Similar pathways of tumorgenesis in both species are a 7ofSMAD4. This, however, was not observed for prerequisite for our cross-species strategy for distin- exon 4 of SMAD4 and the four reference control guishing between cancer drivers and passengers. exons described above. To conclude these observations Besides SMAD4, we also sequenced a few other genes more quantitatively, we performed t-test analyses to in 18q21.1-q21.2 (Figure 2B), including methyl-CpG examine the Ct differences between the tumors and binding MBD1 and MBD2, as controls. While their matching normal samples. We found that the dif- we did not find stop codon or indel mutations in coding ferences were significant for SMAD4 exon7(t=2and exons of the two genes in the mouse tumors, we did p is between 0.05-0.025) but insignificant for the other identify such mutations in exons 5, 6, 8, 13 of MBD1 exonsexamined(t<0.5andp>0.25).Theseresults (but not in any MBD2 exons) in the human tumors (see were consistent with the sequencing analyses, and indi- Additional file 1). These findings are consistent with the cated that SMAD4 exon 7 was likely deleted in the 5q22.2 sequencing analysis described above, which mouse tumors. revealed an overall higher sequence mutation rate in the Because heterogeneity is expected for somatic muta- human tumors than in the mouse tumors, supporting tions, we were puzzled by the apparent homogeneity that our sequencing analysis was run properly as well observed for SMAD4 alteration in the mouse tumors with the 18q21.1-q21.2 genes. (i.e., only exon 7 found to be disrupted). SMAD4 exon 7 SMAD4 was disrupted in mouse tumors is 51bp in size and is separated from the previous exon To determine whether SMAD4 was disrupted or intact by an intron of 1.3 kb and from the following exon by in the mouse tumors, we first sequenced its exons. For another intron of 7.3 kb. Thus, the genomic regions sur- coding exon 7 of mouse SMAD4,wefoundthat,while rounding SMAD4,exon7,amountto8.6kb.These 60% of the matching-normal samples were sequenced regions were found to harbor more L1 s and simple successfully, all tumors failed the sequencing procedure. repeats with a low (< 10%) sequence divergence among However, the remaining 10 coding exons (mouse the copies (i.e., younger repeats) and to be less con- SMAD4 has 11 coding exons with a total of 1,656 bases) served in other species (e.g., a gap of 5.8 kb for mouse were sequenced successfully with all tumor and normal and 4 kb for humans were found when aligning the Ji et al. BMC Cancer 2010, 10:426 Page 9 of 13 http://www.biomedcentral.com/1471-2407/10/426

corresponding sequences of this region), compared with P53 also altered in mouse tumors the rest of the gene (see Additional file 1). We hypothe- Besides genes of 5q22.2 and 18q21.1-q21.2 described size that deletions of exon 7 could have occurred via above, we also investigated genes with high mutation homologous recombination between the two highly prevalence (i.e., P53, KRAS,andPIK3CA), as well as identical AT-rich repeats inside its two flanking introns those with intermediate-low mutation prevalence (i.e., (see Additional file 1). Hence, the apparent homogeneity SMAD2 and SMAD3) in human CRC [46]. This would could have arisen from the fact that sequences around provide more evidence to evaluate the molecular simila- exon 7 are intrinsically more unstable, compared with rities of tumorigenesis between human CRC and the those of other parts of the gene. A somewhat similar ApcMin/+ model. situation has been reported for the BRCA1 gene locus, We conducted qRT-PCR analysis on these genes along of which the structural instability revealed by evolution- with two reference genes b-actin and GAPDH,with12 ary analyses may mostly originate from Alu-mediated matched colon adenomas and normal tissue samples rearrangement [43]. from several ApcMin/+ mice. After normalization with b- Our hypothesis was partially supported by the obser- actin, we performed t-tests to determine if the Ct differ- vation of “subtle” chromosome instability (CIN) in the ence between the tumors and normal samples is signifi- mouse tumor genomes. By analyzing several mouse cant for each gene. We found that P53 was significantly tumors with Roche NimbleGen’s 384 K oligonucleotide altered in these tumors (t = -4.8; p < 0.0001). However, comparative genomic hybridization (CGH) arrays, we we did not detect significant changes for the other found that ~4% of the genomic sequences of the genes (KRAS, PIK3CA, SMAD2,andSMAD3), although tumors were amplified/deleted, compared to 1% when their t-values are >10 times larger than that of GAPDH hybridizing a normal sample against another normal (see Table s1 in Additional file 1). Hence, this analysis sample. This is consistent with a recent study that indicates that P53 was also altered in the mouse tumors, observed misoriented spindles and misaligned chromo- adding another piece of evidence supporting the mole- somes associated with tetraploid genotypes in dividing cular similarities between human CRC and the mouse crypt cells within the small intestines of APCMin/+ model. For the other genes, further analysis (e.g., mice [44]. However, we must emphasize the word sequencing) are clearly needed to determine if they are “subtle” for the observed mouse CIN because, for the intact or disrupted in the mouse tumors. human tumors displaying CIN, the amplified/deleted sequences amount to >10% of the genome. This, per- Discussion haps, explains why other studies did not find CIN in The ability to distinguish between cancer driver and ApcMin/+ tumors [45]. The reason that “subtle” CIN passenger alterations has been a central aim of cancer was detected here is, most likely, due to the high reso- research. To achieve this, we are developing a cross- lution of the oligo arrays being used (one probe every species comparison strategy by taking advantage of the 5-6 kb across the genome), and, as far as we know, we difference in the genomic location of orthologous genes are the only group that has used such high density between the human and other mammals. To showcase arrays on these mouse tumors. this strategy, we conducted a pilot study focusing on With the Ct difference between the tumors and their genes APC and MCC, which are adjacent in 5q22.2 of matching normal samples ranging from -0.1 to 2, the 10 the human genome but distant in the mouse genome tumors used in qPCR analyses showed more heteroge- (Figure 2). We found that, consistent with literature neity than the 5 tumors used in the sequencing analysis reports, both genes were disrupted in human colorectal for SMAD4 exon7 (all tumors failed the sequencing pro- tumors. However, in the mouse tumors, only bona fide cess). One likely reason is that these 5 tumors were con- CRC gene APC was disrupted whereas MCC appeared siderably larger than those used in qPCR analyses (4-5 to be intact. This indicates that MCC alteration may not mm versus 1-2 mm in size). Therefore, they were at be a driver (but rather a passenger) of colorectal tumori- later tumor progression stages, where more extensive genesis, and we hypothesize that its disruption in human genomic instability develops and more genomic CRCs could merely be a result of its close proximity to abnormalities accumulate. APC in the human genome (Figures 2 &4), as rationa- SMAD4 in human tumors lizedasfollows.WhenmutagensactontheAPC locus A high sequencing failure rate was also found in the causing APC mutations, the neighboring MCC gene is human homologue of the mouse SMAD4 coding exon 7. accidently exposed to the mutagens (consistent with There were also disruptions in a number of other cod- this, we also found a significant amount of nonsense ing exons (see Additional file 1), indicating that SMAD4 mutations and indels in another neighboring gene DP1 was also altered in the human tumors, consistent with or deletedinpolyposis1; see Additional file 1). Because previous studies [25-27]. MCC is 10 Mb away from APC in the mouse genome Ji et al. BMC Cancer 2010, 10:426 Page 10 of 13 http://www.biomedcentral.com/1471-2407/10/426

Figure 4 Our study indicates that MCC alteration is unlikely to be a driver of colorectal tumorigenesis at early stages (top). The bottom illustrates our hypothesis that alteration of MCC in human CRCs occurs as a result of its proximity to APC in the human genome (~200 kb apart). However, because MCC is distant from APC (~10 Mb apart) in the mouse genome, agents that caused APC alternations have no effect on MCC; thus, like other genes without causative roles in tumorigenesis, MCC remains intact in the mouse tumors.

(Figure 2), agents that cause APC alteration may essen- is distantly located from the ENU-induced ApcMin/+ tially have no effect on MCC. Hence, MCC could simply locus (Figure 2). This is further supported by our obser- be like other background genes with no causative roles vation that the expression of P53 is also significantly (Figure 4), and thus no enrichment of MCC alteration altered in mouse tumors. Consistent with these, several was observed in the mouse tumors. Perhaps a more recent studies reported that a subset of the genes, with familiar analogous example of this situation would be: a expression altered in ApcMin/+ adenomas, were con- passenger was injured in a car accident that was caused firmed to be changed in human CRCs as well [40,48]. In by the driver of the car (MCC vs. APC in the human addition, our high density aCGH array analysis found tumors); however, if the passenger had been sitting in a subtle CIN in the genome of the mouse adenomas, different car, he/she would, most likely, have been which is consistent with the human colorectal tumori- injury-free (MCC vs. APC in the mouse tumors). genesis model where CIN is an early event (Figure 1). This pilot study demonstrates the promise of using Of course, a significant amount of work is needed to this cross-species comparative strategy to identify cancer conclude that APCMin/+ tumorigenesis indeed share the driver alterations. Nearly 10,000 loci have been found to similar pathway as human CRC as shown in Figure 1, be disrupted in human colonic polyps [47]; clearly, it such as investigating other contributors such as KRAS would be useful to know which disruptions are colorec- gene alteration. tal tumorigenesis-drivers and which ones are not. Sev- Besides conducting further studies with B6 ApcMin/+, eral groups, including us, have built a high resolution other mouse models should also be explored. This is synteny/rearrangement map between the human and especially needed when comparing more genetically mouse genomes [8,9,18], and identified ~400 inter- homogeneous gene-knockout mouse models to more species rearrangements. Thus, there are many places genetically heterogeneous sporadic human CRCs. The like the APC/MCC locusinthehumangenome.Pro- APCMin/+ model may only represent a subset of sporadic videdthattheB6ApcMin/+ model is clearly demon- human CRCs. This could be improved by studying other strated to share similar cancer progression pathways as mouse models, such as SMAD4-knockout mouse models human CRCs (discussed below), expanding this pilot [28,29] and those described in a recent publication project to the entire genome would potentially identify (azoxymethane or AOM, Smad3-/-,andTgfb1-/- Rag2-/-) many targets like MCC, facilitating bona fide CRC gene [48]. Importantly, tumors from APCMin/+ are mostly discovery. adenomas, and obviously APCMin/+cannot represent late The disruption of SMAD4 in mouse tumors raises the stage human CRCs. Other models, such as the one possibility that B6 ApcMin/+ mice might share a similar described by Hinoi et al. [49] that aims to model transi- genetic pathway of tumorigenesis as human CRC shown tion from adenoma to adenocarinoma, should be inves- in Figure 1 [12,13], especially considering that SMAD4 tigated. Of course, sporadic cancers from animal models Ji et al. BMC Cancer 2010, 10:426 Page 11 of 13 http://www.biomedcentral.com/1471-2407/10/426

would be the best representation of sporadic human Additional material cancers [50], but sporadic cancers are rare in mice and difficult to obtain in other species. However, essentially Additional file 1: Supporting Material. This file provides additional an unlimited amount of tumors could be accessed with analyses and information to support the conclusions described in the the genetically modified mouse models, which greatly main text. facilitates the studies. Unlike strategies such as gene knockout mouse mod- els, our approach does not provide direct evidence to Acknowledgements determine whether a gene alteration is a cancer driver We thank Drs. Kelly Moreman and Alison V. Nairn for their help on the qRT- PCR experiments; Ms. Celestia Davis for technical help in tumor isolation or passenger. However, this approach is much more from ApcMin/+ mice; Dr. Wei-Wen Cai for performing BAC aCGH analysis; Ms. cost- and time-effective, as performing large scale ana- Lydia Youmans, Mr. Benjamin Nelson, Mr. Edwin Shin, Ms. Candra Tran, Mr. lyses, such as microarray or sequencing studies, could Chris Rodgers, Mr. Geoffrey Iles for their contribution in qPCR and qRT-PCR analyses; and Ms. Jennifer Slife and Mr. Benjamin Nelson for editing the allow all the genes and the entire genome of an organ- manuscript. This work is funded by the American Cancer Society and ismtobeexaminedinasingleexperiment.Our Georgia Cancer Coalition. approach could quickly narrow down the list of cancer- Author details driver gene alteration candidates, which then can be ver- 1Department of Biochemistry and Molecular Biology, Institute of ified by direct approaches, such as gene knockout mouse Bioinformatics, University of Georgia, Athens 30602, GA, USA. 2McArdle models. Laboratory for Cancer Research, University of Wisconsin Madison, WI 53703, USA. 3J. Craig Venter Institute, Rockville, MD 20850, USA. 4Center for Colon Cancer Research, University of South Carolina, Columbia, SC 29208, USA. Conclusions 5Departments of Surgery, Pathology, and Biostatistics, H. Lee Moffitt Cancer We conducted a pilot study to demonstrate the effec- Center and Research Institute, Tampa, FL 33612, USA. tiveness of a cross-species comparison strategy to distin- Authors’ contributions guish between cancer driver and passenger alteration XJ and JT analyzed the sequence data. RH and MP provided the mouse candidates, by utilizing the difference in the genomic tumors and associated pathological information. DB and SF did the sequencing work. CV and TJ provided the human DNA samples and location of orthologous genes and genomic loci between associated pathological/clinical information. SZ participated in conducting the human and other mammals. This pilot study the qPCR and qRT-PCR analyses, designed the study, and wrote the focused on the human 5q22.2 and 18q21.1-q21.2 manuscript. All authors contributed to the manuscript editing. All authors regions, which are frequently disrupted in human CRCs. read and approved the final manuscript. They also encode bona fide cancer genes (e.g., APC, Competing interests SMAD4), an apparent passenger gene, DCC, and genes The authors declare that they have no competing interests. that are disrupted in human CRCs but whose roles in Received: 31 August 2009 Accepted: 13 August 2010 cancer remain unclear (e.g., MCC). Both regions are Published: 13 August 2010 evolutionarily unstable. As a result, APC and MCC are adjacent in the human genome but are distant in the References 1. Hampton OA, Den Hollander P, Miller CA, Delgado DA, Li J, Coarfa C, mouse genome. By studying the same orthologous genes Harris RA, Richards S, Scherer SE, Muzny DM, Gibbs RA, Lee AV, in colon tumors from humans and the CRC mouse Milosavljevic A: A sequence-level map of chromosomal breakpoints in model B6 ApcMin/+, we found that bona fide cancer the MCF-7 breast cancer cell line yields insights into the evolution of a cancer genome. Genome Res 2009, 19:167-77. genes APC and SMAD4 were disrupted in both species. 2. Pole JC, Courtay-Cahen C, Garcia MJ, Blood KA, Cooke SL, Alsop AE, However, the questionable cancer gene MCC appeared Tse DM, Caldas C, Edwards PA: High-resolution analysis of chromosome to be intact in the mouse tumors, unlike in the human rearrangements on 8p in breast, colon and pancreatic cancer reveals a MCC complex pattern of loss, gain and translocation. Oncogene 2006, tumors. This indicates that may not be a driver of 25:5693-706. colorectal tumorigenesis, and that its alteration in 3. Greenman C, Stephens P, Smith R, Dalgliesh GL, Hunter C, Bignell G, human CRCs could be due to its close proximity to Davies H, Teague J, Butler A, Stevens C, Edkins S, O’Meara S, Vastrik I, Schmidt EE, Avis T, Barthorpe S, Bhamra G, Buck G, Choudhury B, APC in the human genome (Figures 2 &4). This pilot Clements J, Cole J, Dicks E, Forbes S, Gray K, Halliday K, Harrison R, Hills K, study demonstrates the promise of using this cross- Hinton J, Jenkinson A, Jones D, et al: Patterns of somatic mutation in species comparative genomics and oncology strategy to human cancer genomes. Nature 2007, 446:153-8. identify cancer-causative alterations. Once the B6 4. Jones S, Zhang X, Parsons DW, Lin JC, Leary RJ, Angenendt P, Mankoo P, Min/+ Carter H, Kamiyama H, Jimeno A, Hong SM, Fu B, Lin MT, Calhoun ES, Apc model is clearly demonstrated to share similar Kamiyama M, Walter K, Nikolskaya T, Nikolsky Y, Hartigan J, Smith DR, tumorigenesis pathways as in humans CRCs (Figure 1), Hidalgo M, Leach SD, Klein AP, Jaffee EM, Goggins M, Maitra A, Iacobuzio- Donahue C, Eshleman JR, Kern SE, Hruban RH, Karchin R, Papadopoulos N, expanding this pilot project to the entire genome would Parmigiani G, Vogelstein B, Velculescu VE, Kinzler KW: Core signaling potentially identify many targets like MCC, facilitating pathways in human pancreatic cancers revealed by global genomic cancer driver gene alteration discovery. analyses. Science 2008, 321:1801-6. Ji et al. BMC Cancer 2010, 10:426 Page 12 of 13 http://www.biomedcentral.com/1471-2407/10/426

5. Haverty PM, Fridlyand J, Li L, Getz G, Beroukhim R, Lohr S, Wu TD, Cavet G, 26. Thiagalingam S, Lengauer C, Leach FS, Schutte M, Hahn SA, Overhauser J, Zhang Z, Chant J: High-resolution genomic and expression analyses of Willson JK, Markowitz S, Hamilton SR, Kern SE, Kinzler KW, Vogelstein B: copy number alterations in breast tumors. Genes Cancer Evaluation of candidate tumour suppressor genes on chromosome 18 in 2008, 47:530-42. colorectal cancers. Nat Genet 1996, 13:343-6. 6. Beroukhim R, Getz G, Nghiemphu L, Barretina J, Hsueh T, Linhart D, 27. Riggins GJ, Kinzler KW, Vogelstein B, Thiagalingam S: Frequency of Smad Vivanco I, Lee JC, Huang JH, Alexander S, Du J, Kau T, Thomas RK, Shah K, gene mutations in human cancers. Cancer Res 1997, 57:2578-80. Soto H, Perner S, Prensner J, Debiasi RM, Demichelis F, Hatton C, Rubin MA, 28. Takaku K, Miyoshi H, Matsunaga A, Oshima M, Sasaki N, Taketo MM: Gastric Garraway LA, Nelson SF, Liau L, Mischel PS, Cloughesy TF, Meyerson M, and duodenal polyps in Smad4 (Dpc4) knockout mice. Cancer Res 1999, Golub TA, Lander ES, Mellinghoff IK, Sellers WR: Assessing the significance 59:6113-7. of chromosomal aberrations in cancer: Methodology and application to 29. Takaku K, Oshima M, Miyoshi H, Matsui M, Seldin MF, Taketo MM: Intestinal glioma. PNAS 2007, 104:20007-20012. tumorigenesis in compound mutant mice of both Dpc4 (Smad4) and 7. Voidonikolas G, Kreml SS, Chen C, Fisher WE, Brunicardi FC, Gibbs RA, Apc genes. Cell 1998, 92:645-56. Gingras MC: Basic principles and technologies for deciphering the 30. Fearon ER, Cho KR, Nigro JM, Kern SE, Simons JW, Ruppert JM, Hamilton SR, genetic map of cancer. World J Surg 2009, 33:615-29. Preisinger AC, Thomas G, Kinzler KW, Vogelstein B: Identification of a 8. Zhao S, Shetty J, Hou L, Delcher A, Zhu B, Osoegawa K, de Jong P, chromosome 18q gene that is altered in colorectal cancers. Science 1990, Nierman WC, Strausberg RL, Fraser CM: Human, Mouse and Rat Genome 247:49-56. Large Scale Rearrangements: Stability Versus Speciation. Genome Res 31. Fazeli A, Dickinson SL, Hermiston ML, Tighe RV, Steen RG, Small CG, 2004, 14:1851-60. Stoeckli ET, Keino-Masu K, Masu M, Rayburn H, Simons J, Bronson RT, 9. Pevzner P, Tesler G: Human and mouse genomic sequences reveal Gordon JI, Tessier-Lavigne M, Weinberg RA: Phenotype of mice lacking extensive breakpoint reuse in mammalian evolution. PNAS 2003, functional Deleted in colorectal cancer (Dcc) gene. Nature 1999, 100:7672-7. 386:796-804. 10. Moser AR, Luongo C, Gould KA, McNeley MK, Shoemaker AR, Dove WF: 32. Grade M, Ghadimi BM, Varma S, Simon R, Wangsa D, Barenboim- ApcMin: a mouse model for intestinal and mammary tumorigenesis. Eur Stapleton L, Liersch T, Becker H, Ried T, Difilippantonio MJ: Aneuploidy- J Cancer 1995, 31A:1061-4. dependent massive deregulation of the cellular transcriptome and 11. Moser AR, Mattes EM, Dove WF, Lindstrom MJ, Haag JD, Gould MN: apparent divergence of the Wnt/beta-catenin signaling pathway in ApcMin, a mutation in the murine Apc gene, predisposes to mammary human rectal carcinomas. Cancer Res 2006, 66:267-82. carcinomas and focal alveolar hyperplasias. PNAS 1993, 90:8977-81. 33. Peifer M, Polakis P: Wnt signaling in oncogenesis and embryogenesis–a 12. Kinzler KW, Vogelstein B: Lessons from hereditary colorectal cancer. Cell look outside the nucleus. Science 2000, 287:1606-9. 1996, 87:159-170. 34. Bienz M, Clevers H: Linking colorectal cancer to Wnt signaling. Cell 2000, 13. Rajagopalan H, Nowak MA, Vogelstein B, Lengauer C: The significance of 103:311-20. unstable chromosomes in colorectal cancer. Nat Rev Cancer 2003, 35. Clements WM, Lowy AM, Groden J: Adenomatous polyposis coli/beta- 3:695-701. catenin interaction and downstream targets: altered gene expression in 14. Habermann JK, Paulsen U, Roblick UJ, Upender MB, McShane LM, Korn EL, gastrointestinal tumors. Clin Colorectal Cancer 2003, 3:113-20. Wangsa D, Krüger S, Duchrow M, Bruch HP, Auer G, Ried T: Stage-specific 36. Fearnhead NS, Britton MP, Bodmer WF: The ABC of APC. Hum Mol Genet alterations of the genome, transcriptome, and proteome during 2001, 10:721-33. colorectal carcinogenesis. Genes Chromosomes Cancer 2007, 46:10-26. 37. Clarke AR: Studying the consequences of immediate loss of gene 15. Camps J, Grade M, Nguyen QT, Hörmann P, Becker S, Hummon AB, function in the intestine: APC. Biochem Soc Trans 2005, 33:665-6. Rodriguez V, Chandrasekharappa S, Chen Y, Difilippantonio MJ, Becker H, 38. Kinzler KW, Nilbert MC, Vogelstein B, Bryan TM, Levy DB, Smith KJ, Ghadimi BM, Ried T: Chromosomal breakpoints in primary colon cancer Preisinger AC, Hamilton SR, Hedge P, Markham A, Carlson M, Joslyn G, cluster at sites of structural variants in the genome. Cancer Res 2008, Groden J, White R, Miki Y, Miyoshi Y, Nishisho I, Nakamura Y: Identification 68:1284-95. of a gene located at chromosome 5q21 that is mutated in colorectal 16. Camps J, Armengol G, del Rey J, Lozano JJ, Vauhkonen H, Prat E, Egozcue J, cancers. Science 1991, 251:1366-70. Sumoy L, Knuutila S, Miró R: Genome-wide differences between 39. Boivin GP, Groden J: Mouse Models of Intestinal Cancer. Comp Med 2004, microsatellite stable and unstable colorectal tumors. Carcinogenesis 2006, 54:15-8. 27:419-2. 40. Reichling T, Goss KH, Carson DJ: Transcriptional profiles of intestinal 17. Toft NJ, Arends MJ: DNA mismatch repair and colorectal cancer. J Pathol tumors in Apc(Min) mice are unique from those of embryonic intestine 1998, 185:123-129. and identify novel gene targets dysregulated in human colorectal 18. Ji X, Zhao S: DA and Xiao-two giant and composite LTR-retrotransposon- tumors. Cancer Res 2005, 65:166-76. like elements identified in the human genome. Genomics 2008, 91:249-58. 41. Shoemaker AR, Moser AR, Midgley CA, Clipson L, Newton MA, Dove WF: A 19. Askew DJ, Askew YS, Kato Y, Turner RF, Dewar K, Lehoczky J, Silverman GA: resistant genetic background leading to incomplete penetrance of Comparative genomic analysis of the clade B serpin cluster at human intestinal neoplasia and reduced loss of heterozygosity in ApcMin/+ mice. chromosome 18q21: amplification within the mouse squamous cell PNAS 1998, 95:10826-31. carcinoma antigen gene locus. Genomics 2004, 84:176-84. 42. Haigis KM, Hoff PD, White A, Shoemaker AR, Halberg RB, Dove WF: Tumor 20. Deininger PL, Batzer MA: Mammalian retroelements. Genome Res 2002, regionality in the mouse intestine reflects the mechanism of loss of Apc 12:1455-65. function. PNAS 2004, 101:9769-73. 21. Venter JC, Adams MD, Myers EW, Li PW, Mural RJ, Sutton GG, Smith HO, 43. Pavlicek A, Noskov VN, Kouprina N, Barrett JC, Jurka J, Larionov V: Evolution Yandell M, Evans CA, Holt RA, Gocayne JD, Amanatides P, Ballew RM, of the tumor suppressor BRCA1 locus in primates: implications for Huson DH, Wortman JR, Zhang Q, Kodira CD, Zheng XH, Chen L, Skupski M, cancer predisposition. Hum Mol Genet 2004, 13:2737-51. Subramanian G, Thomas PD, Zhang J, Gabor Miklos GL, Nelson C, Broder S, 44. Caldwell CM, Green RA, Kaplan KB: APC mutations lead to cytokinetic Clark AG, Nadeau J, McKusick VA, Zinder N, et al: The Sequence of the failures in vitro and tetraploid genotypes in Min mice. J Cell Biol 2007, Human Genome. Science 2001, 291:1304-1351. 178:1109-20. 22. International Human Genome Sequencing Consortium: Initial sequencing 45. Halberg RB, Waggoner J, Rasmussen K, White A, Clipson L, Prunuske AJ, and analysis of the human genome. Nature 2001, 409:860-921. Bacher JW, Sullivan R, Washington MK, Pitot HC, Petrini JH, Albertson DG, 23. Shi Y, Massague J: Mechanisms of TGF-beta signaling from cell Dove WF: Long-lived Min mice develop advanced intestinal cancers membrane to the nucleus. Cell 2003, 113:685-700. through a genetically conservative pathway. Cancer Res 2009, 69:5768-75. 24. Massague J, Blain SW, Lo RS: TGFbeta signaling in growth control, cancer, 46. Sjöblom T, Jones S, Wood LD, Parsons DW, Lin J, Barber TD, Mandelker D, and heritable disorders. Cell 2000, 103:295-309. Leary RJ, Ptak J, Silliman N, Szabo S, Buckhaults P, Farrell C, Meeh P, 25. Hahn SA, Schutte M, Hoque AT, Moskaluk CA, da Costa LT, Rozenblum E, Markowitz SD, Willis J, Dawson D, Willson JK, Gazdar AF, Hartigan J, Wu L, Weinstein CL, Fischer A, Yeo CJ, Hruban RH, Kern SE: DPC4, a candidate Liu C, Parmigiani G, Park BH, Bachman KE, Papadopoulos N, Vogelstein B, at human chromosome 18q21.1. Science 1996, Kinzler KW, Velculescu VE: The consensus coding sequences of human 271:350-3. breast and colorectal cancers. Science 2006, 314:268-74. Ji et al. BMC Cancer 2010, 10:426 Page 13 of 13 http://www.biomedcentral.com/1471-2407/10/426

47. Stoler DL, Chen N, Basik M, Kahlenberg MS, Rodriguez-Bigas MA, Petrelli NJ, Anderson GR: The onset and extent of genomic instability in sporadic colorectal tumor progression. PNAS 1999, 96:15121-6. 48. Kaiser S, Park YK, Franklin JL, Halberg RB, Yu M, Jessen WJ, Freudenberg J, Chen X, Haigis K, Jegga AG, Kong S, Sakthivel B, Xu H, Reichling T, Azhar M, Boivin GP, Roberts RB, Bissahoyo AC, Gonzales F, Bloom GC, Eschrich S, Carter SL, Aronow JE, Kleimeyer J, Kleimeyer M, Ramaswamy V, Settle SH, Boone B, Levy S, Graff JM, Doetschman T, Groden J, Dove WF, Threadgill DW, Yeatman TJ, Coffey RJ Jr, Aronow BJ: Transcriptional recapitulation and subversion of embryonic colon development by mouse colon tumor models and human colon cancer. Genome Biol 2007, 8:R131. 49. Hinoi T, Akyol A, Theisen BK, Ferguson DO, Greenson JK, Williams BO, Cho KR, Fearon ER: Mouse model of colonic adenoma-carcinoma progression based on somatic Apc inactivation. Cancer Res 2007, 67:9721-30. 50. Tang J, Le S, Sun L, Yan X, Zhang M, Macleod J, Leroy B, Northrup N, Ellis A, Yeatman TJ, Liang Y, Zwick ME, Zhao S: Copy number abnormalities in sporadic canine colorectal cancers. Genome Res 2010, 20:341-50.

Pre-publication history The pre-publication history for this paper can be accessed here: http://www.biomedcentral.com/1471-2407/10/426/prepub

doi:10.1186/1471-2407-10-426 Cite this article as: Ji et al.: Distinguishing between cancer driver and passenger gene alteration candidates via cross-species comparison: a pilot study. BMC Cancer 2010 10:426.

Submit your next manuscript to BioMed Central and take full advantage of:

• Convenient online submission • Thorough peer review • No space constraints or color figure charges • Immediate publication on acceptance • Inclusion in PubMed, CAS, Scopus and Google Scholar • Research which is freely available for redistribution

Submit your manuscript at www.biomedcentral.com/submit